采用改进黑猩猩优化算法的特征选择

Feature Selection Based on Improved Chimp Optimization Algorithm

针对黑猩猩优化算法在勘探阶段可能会陷入局部最优的缺点,提出了一种改进的黑猩猩优化算法。首先计算适应度值最高和最低的黑猩猩之间的斯皮尔曼等级相关系数,然后采用甲虫天线搜索算法使得适应度值低的黑猩猩获得嗅觉能力,以提高适应度值低的黑猩猩的局部勘探开发能力。同时为解决传统支持向量机在封装式特征选择中分类精度较差的问题,将改进后的黑猩猩优化算法与支持向量机相结合,实现同步优化。为了验证所提方法的有效性,选取了UCI(University of California,Irvine)数据库中的10个数据集进行实验,从精确度、所用特征个数和适应度值三个方面与常用算法进行比较和综合评价,实验结果表明该算法具有更好的精确度和稳定性,有效适用于特征选择这一工程问题。

Aiming at the shortcoming that the chimp optimization algorithm(CHOA) may fall into local optimum in the exploration stage, an improved chimp optimization algorithm(ICHOA) is proposed. Firstly, the Spearman rank correlation coefficient between the chimps with the highest fitness value and the chimps with the lowest fitness value is calculated, and then the beetle antenna search algorithm(BAS) is used to make the chimps with low fitness value obtain the olfactory ability,so as to improve the local exploration and development ability of chimps with low fitness value. Meantime, in order to solve the problem of poor classification accuracy of traditional support vector machine in wrapper feature selection, the ICHOA is combined with support vector machine to realize synchronous optimization. To verify the effectiveness of the proposed method, 10 datasets in UCI database are selected for experiments. The algorithm is compared with the common algorithms in terms of accuracy, the number of features used and the fitness value. The experimental results show that the algorithm has better accuracy and stability, which can be applied to the engineering problem of feature selection effectively.